Unpredictable growth of knowledge

16 views
Skip to first unread message

Dennis C. Hackethal

unread,
Mar 29, 2019, 3:49:35 PM3/29/19
to fallibl...@googlegroups.com
I was struggling the other day to explain to someone why the growth of knowledge is inherently unpredictable. I *think* I can explain it in terms of “it’s a generic algorithm, and a genetic algorithm has unpredictable output”, but unless the other party is already familiar with the concept of knowledge being the result of a genetic algorithm, that doesn’t go very far. It also made me think that a genetic algorithm is unpredictable *to a degree*. If someone runs a genetic algorithm for eg the traveling salesman problem, they know it’s going to return a solution to the problem in terms of distances etc, and not something completely unexpected. So there’s at least some way to constrain the space of possible answers. I don’t think it’s possible to constrain human answers in this way, but I don’t think I understand why. I also don’t know if probabilistic = unpredictable (my guess is “no”).

Alan Forrester

unread,
Mar 30, 2019, 7:06:32 AM3/30/19
to FIGG
On 29 Mar 2019, at 19:49, 'Dennis C. Hackethal' via Fallible Ideas <fallibl...@googlegroups.com> wrote:

> I was struggling the other day to explain to someone why the growth of knowledge is inherently unpredictable. I *think* I can explain it in terms of “it’s a generic algorithm, and a genetic algorithm has unpredictable output”, but unless the other party is already familiar with the concept of knowledge being the result of a genetic algorithm, that doesn’t go very far. It also made me think that a genetic algorithm is unpredictable *to a degree*. If someone runs a genetic algorithm for eg the traveling salesman problem, they know it’s going to return a solution to the problem in terms of distances etc, and not something completely unexpected. So there’s at least some way to constrain the space of possible answers. I don’t think it’s possible to constrain human answers in this way, but I don’t think I understand why. I also don’t know if probabilistic = unpredictable (my guess is “no”).

If you know that tomorrow you will know X then you would already know X. So you should start working on some other problem that X doesn’t solve.

Also, let’s suppose X is some existing piece of knowledge, like cooking a Dundee cake:

https://www.deliaonline.com/recipes/meals-and-courses/afternoon-tea/traditional-dundee-cake

Since you don’t know how to cook a Dundee cake, you don’t know how you’re going to act on your Dundee cake knowledge. You might not even learn how to cook a Dundee cake. You might look at the recipe and see that it recommends some series of steps that would take a long time. Or you might read that Dundee cake tastes better if you leave it in airtight container and decide you don’t want to wait for cake that long. So then you might decide to postpone or cancel your attempt to learn how to cook a Dundee cake. You might decide to change the ingredients of a traditional Dundee cake slightly to use different dried fruit. So you don’t know all of the details of what you will learn even if you know some outline of what you’ve decided to learn.

Also, creating new knowledge can involve creating knowledge about what priorities you should have. If you think X is worth knowing today, you might think you’ll know X tomorrow. But while you’re learning X you might decide it’s not worth knowing so tomorrow you won’t know X.

Alan

anonymous FI

unread,
Mar 30, 2019, 1:35:19 PM3/30/19
to 'Dennis C. Hackethal' via Fallible Ideas

On Mar 29, 2019, at 12:49 PM, 'Dennis C. Hackethal' via Fallible Ideas
Try reviewing the argument given in BoI.

Dennis C. Hackethal

unread,
Apr 1, 2019, 6:52:41 PM4/1/19
to fallibl...@googlegroups.com
On Sat, Mar 30, 2019 at 4:06 AM 'Alan Forrester' via Fallible Ideas
<fallibl...@googlegroups.com> wrote:
>
> On 29 Mar 2019, at 19:49, 'Dennis C. Hackethal' via Fallible Ideas <fallibl...@googlegroups.com> wrote:
>
> > I was struggling the other day to explain to someone why the growth of knowledge is inherently unpredictable. I *think* I can explain it in terms of “it’s a generic algorithm, and a genetic algorithm has unpredictable output”, but unless the other party is already familiar with the concept of knowledge being the result of a genetic algorithm, that doesn’t go very far. It also made me think that a genetic algorithm is unpredictable *to a degree*. If someone runs a genetic algorithm for eg the traveling salesman problem, they know it’s going to return a solution to the problem in terms of distances etc, and not something completely unexpected. So there’s at least some way to constrain the space of possible answers. I don’t think it’s possible to constrain human answers in this way, but I don’t think I understand why. I also don’t know if probabilistic = unpredictable (my guess is “no”).
>
> If you know that tomorrow you will know X then you would already know X. So you should start working on some other problem that X doesn’t solve.

That is a beautifully concise reductio ad absurdum. Thanks, Alan.

anonymous FI

unread,
Apr 1, 2019, 7:15:49 PM4/1/19
to 'Dennis C. Hackethal' via Fallible Ideas

On Apr 1, 2019, at 3:52 PM, 'Dennis C. Hackethal' via Fallible Ideas
<fallibl...@googlegroups.com> wrote:

> On Sat, Mar 30, 2019 at 4:06 AM 'Alan Forrester' via Fallible Ideas
> <fallibl...@googlegroups.com> wrote:
>>
>> On 29 Mar 2019, at 19:49, 'Dennis C. Hackethal' via Fallible Ideas
>> <fallibl...@googlegroups.com> wrote:
>>
>>> I was struggling the other day to explain to someone why the growth
>>> of knowledge is inherently unpredictable. I *think* I can explain it
>>> in terms of “it’s a generic algorithm, and a genetic algorithm
>>> has unpredictable output”, but unless the other party is already
>>> familiar with the concept of knowledge being the result of a genetic
>>> algorithm, that doesn’t go very far. It also made me think that a
>>> genetic algorithm is unpredictable *to a degree*. If someone runs a
>>> genetic algorithm for eg the traveling salesman problem, they know
>>> it’s going to return a solution to the problem in terms of
>>> distances etc, and not something completely unexpected. So there’s
>>> at least some way to constrain the space of possible answers. I
>>> don’t think it’s possible to constrain human answers in this
>>> way, but I don’t think I understand why. I also don’t know if
>>> probabilistic = unpredictable (my guess is “no”).
>>
>> If you know that tomorrow you will know X then you would already know
>> X. So you should start working on some other problem that X doesn’t
>> solve.
>
> That is a beautifully concise reductio ad absurdum. Thanks, Alan.

How did you determine that this is a reductio ad absurdum?

Dennis C. Hackethal

unread,
Apr 2, 2019, 10:48:06 PM4/2/19
to fallibl...@googlegroups.com
IIRC, a reductio ad absurdum works by tentatively assuming the
opposite of an assumption and showing that it leads to some
impossibility, meaning the opposite cannot be true. This seems to me
what Alan did here:

> If you know that tomorrow you will know X
^ this is the tentative assumption of the opposite of the assumption
that the growth of knowledge is unpredictable

> then you would already know X.
^ this shows the impossibility of that assumption.

>
> --
> You received this message because you are subscribed to the Google Groups "Fallible Ideas" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to fallible-idea...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/fallible-ideas/ADD9654F-A97C-4EC2-ABC9-5D774F5DC644%40gmail.com.
> For more options, visit https://groups.google.com/d/optout.

anonymous FI

unread,
Apr 2, 2019, 11:24:27 PM4/2/19
to 'Dennis C. Hackethal' via Fallible Ideas

On Apr 2, 2019, at 7:47 PM, 'Dennis C. Hackethal' via Fallible Ideas
The claim that tomorrow I will know X does not contradict the claim that
I know X today. There's no impossibility there.

Dennis C. Hackethal

unread,
Apr 2, 2019, 11:28:08 PM4/2/19
to fallibl...@googlegroups.com
On Tue, Apr 2, 2019 at 8:24 PM anonymous FI
It depends on how you read "X". I read it as an explanation of
something. So not as in "tomorrow I will know how to build a
computer", but "tomorrow I will know that in order to build a
computer, I need to... [full explanation of how computers work here]".

With that reading, I cannot form the sentence "Tomorrow I will know X"
without already knowing X today.

anonymous FI

unread,
Apr 2, 2019, 11:34:08 PM4/2/19
to 'Dennis C. Hackethal' via Fallible Ideas

On Apr 2, 2019, at 8:27 PM, 'Dennis C. Hackethal' via Fallible Ideas
There is no contradiction between knowing something today and tomorrow.
I can know it on both days. That is common and does not depend on how
you read "X".

You seem to be in hair-splitting, rule-lawyer mode. But you also have
some unstated assumptions or skipped steps in your head, and are being
blind to the obvious. You don't seem to be focusing on the quotes you
quoted:

> If you know that tomorrow you will know X

and

> then you would already know X.

do not create any kind of contradiction or impossibility. You can
already know X and know it tomorrow too.

Dennis C. Hackethal

unread,
Apr 2, 2019, 11:56:10 PM4/2/19
to fallibl...@googlegroups.com
On Tue, Apr 2, 2019 at 8:34 PM anonymous FI
It's certainly possible I misunderstood Alan and it was not intended
as a reductio. If so, perhaps he can shed some light on it.

You can absolutely know something tomorrow and not today. And you can
know it both tomorrow and today. You can even know it today and not
tomorrow, if you forget.

But given I do not know an explanation X today, trying to predict I
will know X tomorrow is impossible, because I would need to know X
today in order to make that prediction. But I already stated that I do
not know X today. Hence the reductio, in my mind.

anonymous FI

unread,
Apr 3, 2019, 12:07:21 AM4/3/19
to 'Dennis C. Hackethal' via Fallible Ideas

On Apr 2, 2019, at 8:55 PM, 'Dennis C. Hackethal' via Fallible Ideas
In this discussion, I have not made a claim about what type of argument
Alan intended. That is not the thing at issue.


> You can absolutely know something tomorrow and not today. And you can
> know it both tomorrow and today. You can even know it today and not
> tomorrow, if you forget.
>
> But given I do not know an explanation X today, trying to predict I
> will know X tomorrow is impossible, because I would need to know X
> today in order to make that prediction. But I already stated that I do
> not know X today. Hence the reductio, in my mind.

You don't seem to be understanding.

You made a claim that it was a reductio ad absurdum. When asked for
details, you said that two quotes together make an impossibility. But
those two quotes are compatible. So you're mistaken.

It seems like your thinking is fuzzy and unable to deal with literal
details or correctness. Instead, you seem to have in mind a bunch of
things you haven't said – perhaps haven't even verbalized to yourself
– and not to realize they're needed for your claims to make sense.
You haven't actually said what your real reasoning is, so we haven't yet
been able to evaluate it. I'll we've done so far is I've tried to
explain that the reasoning you presented is trivially false. I hope this
can lead to considering what your actual reasoning is.

Alan Forrester

unread,
Apr 3, 2019, 3:32:22 AM4/3/19
to FIGG, FI
Dennis quoted the following argument

> If you know that tomorrow you will know X then you would already
> know X. So you should start working on some other problem that X doesn’t
> solve.


and then stated it was a reductio ad absurdum. The quote was poorly chosen because it makes a statement and then a recommendation and omits almost all of the explanation in the post.

Is the argument a reductio ad absurdum. A reductio ad absurdum involves drawing an absurd conclusion from an idea you want to refute:

https://www.iep.utm.edu/reductio/

> Reductio ad absurdum is a mode of argumentation that seeks to establish a contention by deriving an absurdity from its denial, thus arguing that a thesis must be accepted because its rejection would be untenable.


I don’t think reductio ad absurdum is a good way to understand arguments.

One problem is that different people regard different ideas as absurd. Some people regard the rejection of inductivism as absurd and reject Popper’s ideas for that reason. Popperians might say the idea that experiments prove a theory is absurd. So discussing absurdity doesn’t help solve that disagreement.

I didn’t intend my argument as a reductio ad absurdum. The argument is an explanation of why it’s impossible to predict the growth of knowledge.

Alan
Reply all
Reply to author
Forward
0 new messages